Is compatibility with InfluxDB 1.7.x broken?

[M0rdecay]

After upgrading to Grafana 7.0, panels using Flux stopped working. The error is as follows: In this case, the data source is ok: This is a big problem, because InfluxDB newest than 1.7.3 cannot be used in highly loaded environments due to this error - https://github.com/influxdata/influxdb/issues/13318

Are you planning to return compatibility with versions 1.7.x?

[ryantxu]

This plugin is now using this client: https://github.com/influxdata/influxdb-client-go/

I'm not sure what is involved in getting that client to read flux results from 1.7...

[ryantxu]

Any more details on the error you see? Is there anything in your browser console log? or maybe in the server logs?

[M0rdecay]

Here is what information I managed to collect

In browser console:

From "Network" tab:

{results: {A: {error: ": ", refId: "A", series: null, tables: null, dataframes: null}}, message: ": "}
message: ": "
results: {A: {error: ": ", refId: "A", series: null, tables: null, dataframes: null}}
A: {error: ": ", refId: "A", series: null, tables: null, dataframes: null}
dataframes: null
error: ": "
refId: "A"
series: null
tables: null

InfluxDB logs (v1.7.3):

[httpd] 172.17.0.1 - - [20/May/2020:06:10:35 +0000] "POST /api/v2/query?org= HTTP/1.1" 500 95 "-" "influxdb-client-go/1.1.0  (linux; amd64)" 9e520a78-9a60-11ea-8dab-0242ac110002 7944
ts=2020-05-20T06:10:35.201599Z lvl=error msg="[500] - \"failed to compile query: type error 2:6-2:19: missing required parameter \\\"start\\\"\"" log_id=0Km0hH8W000 service=httpd

The time window in the request is set by the range function:

  |> range($range)

Please let me know if you need anything else.

[Akira74]

Hi I also had the same error. This happens when you have setting auth-enabled = true In this case you must enter{username}:{password}into the token field. There will be an error shown 500 Internal Server Error: {"error":"panic: column retentionPeriod:int is not of type string"} , when you save it but you can ignore that. The data source worked again.

[M0rdecay]

In my case, the parameter is auth-enabled = false. I use Influxdb in container and the default valse is false.

List of what I redefine:

    INFLUXDB_DATA_MAX_SERIES_PER_DATABASE: "0"
    INFLUXDB_DATA_MAX_VALUES_PER_TAG: "0"
    INFLUXDB_RETENTION_ENABLED: "true"
    INFLUXDB_RETENTION_CHECK_INTERVAL: "10m"
    INFLUXDB_HTTP_MAX_BODY_SIZE: "50000000"
    INFLUXDB_META_RETENTION_AUTOCREATE: "false"
    INFLUXDB_DATA_INDEX_VERSION: "tsi1"
    INFLUXDB_DATA_WAL_FSYNC_DELAY: "100ms"
    INFLUXDB_HTTP_FLUX_ENABLED: "true"

[ext452]

@Akira74 not working (

[M0rdecay]

Hello! Is there any news on this and https://github.com/grafana/influxdb-flux-datasource/issues/101?

[ryantxu]

@M0rdecay for your time filter, can you try:

|> range(start: v.timeRangeStart, stop:v.timeRangeStop)

rather than $range? if that change works, I'll look at adding some of the old macro expansions to help with migratoin. In v7, we aimed for getting copy/paste from chronograph to work

[M0rdecay]

Thank you, I will try this solution: |> range(start: v.timeRangeStart, stop:v.timeRangeStop)

we aimed for getting copy/paste from chronograph to work

IMHO, this breaks uniformity with Grafana as a whole. Variables are expected to have the same syntax everywhere.

[ryantxu]

Let me know if it works!

I wish variables had the same syntax everywhere! weirdly there are two concepts "template variables" and "macros" -- they look the same, but are actually totally differnt, when this switched to a backend plugin, it switched form a frontend template variable to a backend macro

[M0rdecay]

Hm... It works good with simple request:

from(bucket:"$database/$ret_policy")
  |> range(start: v.timeRangeStart, stop:v.timeRangeStop)
  |> filter(fn: (r) =>
    r._measurement == "cpu" and
    r._field == "usage_system" and
    r.cpu == "cpu-total"
  )

But it fails when request has group() function (Grafana log, 7.0.1, image grafana/grafana:7.0.1):

t=2020-06-01T06:24:30+0000 lvl=eror msg="Metric request error" logger=context userId=1 orgId=1 uname=admin error="rpc error: code = Unavailable desc = transport is closing" remote_addr=xxx.xxx.xxx.xxx
t=2020-06-01T06:24:30+0000 lvl=eror msg="Request Completed" logger=context userId=1 orgId=1 uname=admin method=POST path=/api/ds/query status=500 remote_addr=xxx.xxx.xxx.xxx time_ms=1752 size=34 referer="http://grafana:3001/d/NtNnqDjWk/area?orgId=1&refresh=1m&editPanel=49"
t=2020-06-01T06:24:31+0000 lvl=info msg=Profiler logger=plugins.backend pluginId=grafana-influxdb-flux-datasource enabled=false
t=2020-06-01T06:24:31+0000 lvl=eror msg="Metric request error" logger=context userId=1 orgId=1 uname=admin error="rpc error: code = Unavailable desc = connection error: desc = \"transport: error while dialing: dial unix /tmp/plugin105981337: connect: connection refused\"" remote_addr=xxx.xxx.xxx.xxx
t=2020-06-01T06:24:31+0000 lvl=eror msg="Request Completed" logger=context userId=1 orgId=1 uname=admin method=POST path=/api/ds/query status=500 remote_addr=xxx.xxx.xxx.xxx time_ms=0 size=34 referer="http://grafana:3001/d/NtNnqDjWk/area?orgId=1&refresh=1m&editPanel=49"

Very strange, but everything is fine in the InfluxDB log:

[httpd] 172.17.0.1 - - [01/Jun/2020:06:24:30 +0000] "POST /api/v2/query?org= HTTP/1.1" 200 948 "-" "influxdb-client-go/1.1.0  (linux; amd64)" aec707f3-a3d1-11ea-8d6e-0242ac110002 1596765

Upd. group statement: |> group(columns: ["appl"], mode:"by")

[M0rdecay]

It seems that the situation is also happening on other versions of InfluxDB - #109

[M0rdecay]

This is an example of a query that has a problem. The first filter is intentionally replaced by neutral data, but still displays the essence of the situation:

errs = from(bucket: "$database/$ret_policy")
  |> range(start: v.timeRangeStart, stop:v.timeRangeStop)
  |> filter(fn:(r) => (r._measurement == "cpu" 
                       or r._measurement == "mem")
                   and (r._field == "used_percent"
                       or r._field == "usage_system"))
  |> group(columns: ["host"], mode:"by")
  |> derivative(unit: $coll_interval, nonNegative: true)
  |> aggregateWindow(every: $coll_interval, fn: sum)
  |> filter(fn: (r) => r["_value"] >= 0)
  |> yield()

But this request is working fine:

errs = from(bucket: "$database/$ret_policy")
  |> range(start: v.timeRangeStart, stop:v.timeRangeStop)
  |> filter(fn:(r) => (r._measurement == "cpu" 
                       or r._measurement == "mem")
                   and (r._field == "used_percent"
                       or r._field == "usage_system"))

  |> derivative(unit: $coll_interval, nonNegative: true)
  |> aggregateWindow(every: $coll_interval, fn: sum)
  |> filter(fn: (r) => r["_value"] >= 0)
  |> yield()

[M0rdecay]

Let's leave this issue open with the following wording - for compatibility with previous versions of the plugin needs support for the $range variable.

A discussion of the function of the group() function, let's continue at #105

[andig]

for compatibility with previous versions of the plugin needs support for the $range variable.

How are dashboard variables handled? Are those still supported with their $var notation?

[M0rdecay]

Are those still supported with their $var notation?

Yes, the rest of the variables work, I had no problems using them. Only plugin-specific variables have changed. The full list can be found here - https://github.com/grafana/influxdb-flux-datasource/blob/master/pkg/influx/macros.go

This plugin really needs documentation!